import warnings
warnings.filterwarnings('ignore')
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style("whitegrid")
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score, explained_variance_score
import joblib
df = pd.read_csv("Datasets/plant_growth_dataset.csv")
df.head()
| plant_species | date | plant_age_days | temperature | humidity | soil_moisture | soil_ph | light_intensity | height_cm | leaf_area_cm2 | health_score | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | Arive-Dantu | 2023-01-01 | 10 | 24.354431 | 52.994030 | 36.632859 | 7.046809 | 15589.751987 | 9.361922 | 26.532273 | 0.738861 |
| 1 | Arive-Dantu | 2023-01-02 | 10 | 23.921513 | 53.483883 | 37.028936 | 6.960871 | 15683.670310 | 9.371922 | 30.669019 | 0.706514 |
| 2 | Arive-Dantu | 2023-01-03 | 16 | 23.548990 | 52.616118 | 38.361623 | 7.015171 | 15798.653790 | 9.381922 | 33.182690 | 0.668551 |
| 3 | Arive-Dantu | 2023-01-04 | 20 | 23.818069 | 52.247005 | 39.367779 | 7.057316 | 16366.642237 | 9.391922 | 28.648229 | 0.641078 |
| 4 | Arive-Dantu | 2023-01-05 | 16 | 23.891465 | 54.646367 | 38.633777 | 7.012412 | 16514.012045 | 9.401922 | 29.153371 | 0.605280 |
print("Null counts per column:\n", df.isnull().sum())
print("\nTotal null entries:", df.isnull().sum().sum())
# Drop rows with missing values (paper works with small clean dataset).
df = df.dropna().reset_index(drop=True)
# Remove duplicates
dups = df.duplicated().sum()
print("Duplicates found:", dups)
df = df.drop_duplicates().reset_index(drop=True)
print("Shape after dropna & dedupe:", df.shape)
Null counts per column: plant_species 0 date 0 plant_age_days 0 temperature 0 humidity 0 soil_moisture 0 soil_ph 0 light_intensity 0 height_cm 0 leaf_area_cm2 0 health_score 0 dtype: int64 Total null entries: 0 Duplicates found: 0 Shape after dropna & dedupe: (14992, 11)
df.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 14992 entries, 0 to 14991 Data columns (total 11 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 plant_species 14992 non-null object 1 date 14992 non-null object 2 plant_age_days 14992 non-null int64 3 temperature 14992 non-null float64 4 humidity 14992 non-null float64 5 soil_moisture 14992 non-null float64 6 soil_ph 14992 non-null float64 7 light_intensity 14992 non-null float64 8 height_cm 14992 non-null float64 9 leaf_area_cm2 14992 non-null float64 10 health_score 14992 non-null float64 dtypes: float64(8), int64(1), object(2) memory usage: 1.3+ MB
targets = df[["height_cm", "leaf_area_cm2", "health_score"]]
for col in targets.columns:
plt.figure(figsize=(7,4))
sns.histplot(df[col], kde=True, color="green")
plt.title(f"Distribution of {col}")
plt.xlabel(col)
plt.tight_layout()
plt.show()
plt.figure(figsize=(10,6))
sns.pairplot(df[[
"plant_age_days", "temperature", "humidity",
"soil_moisture", "soil_ph", "light_intensity"
]], diag_kind="kde")
plt.show()
<Figure size 1000x600 with 0 Axes>
plt.figure(figsize=(10,6))
sns.heatmap(df.corr(), annot=True, cmap="coolwarm")
plt.title("Feature Correlation")
plt.show()
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
df["plant_species"] = le.fit_transform(df["plant_species"])
import joblib
joblib.dump(le, "Models/label_encoder.sav")
['Models/label_encoder.sav']
X = df[[
"plant_species", "plant_age_days",
"temperature", "humidity", "soil_moisture", "soil_ph", "light_intensity"
]]
y = df[["height_cm", "leaf_area_cm2", "health_score"]]
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, random_state=200)
model_results = {}
def regression_metrics(y_true, y_pred, verbose=True):
results = {}
for i, col in enumerate(y_true.columns):
yt = y_true[col].values
yp = y_pred[:, i]
evar = explained_variance_score(yt, yp)
mae = mean_absolute_error(yt, yp)
mse = mean_squared_error(yt, yp)
rmse = np.sqrt(mse)
r2 = r2_score(yt, yp)
results[col] = {
"EVAR": evar,
"MAE": mae,
"RMSE": rmse,
"R2": r2
}
if verbose:
print(f"\n===== {col} =====")
print(f"EVAR: {evar:.4f}")
print(f"MAE: {mae:.4f}")
print(f"RMSE: {rmse:.4f}")
print(f"R2: {r2:.4f}")
return results
def plot_actual_vs_predicted(y_true, y_pred, title_prefix="Actual vs Predicted"):
for i, col in enumerate(y_true.columns):
plt.figure(figsize=(7, 6))
plt.scatter(y_true[col], y_pred[:, i], alpha=0.6)
mn = min(y_true[col].min(), y_pred[:, i].min())
mx = max(y_true[col].max(), y_pred[:, i].max())
plt.plot([mn, mx], [mn, mx], "r--", lw=2)
plt.xlabel("Actual")
plt.ylabel("Predicted")
plt.title(f"{title_prefix} - {col}")
plt.grid(True)
plt.tight_layout()
plt.show()
from xgboost import XGBRegressor
xgb = XGBRegressor(
random_state=42,
n_estimators=300,
learning_rate=0.05,
max_depth=6,
subsample=0.9,
colsample_bytree=0.9,
n_jobs=-1,
verbosity=0,
tree_method="hist"
)
xgb.fit(X_train, y_train)
# Predict
y_pred_xgb = xgb.predict(X_test)
metrics_xgb = regression_metrics(y_test, y_pred_xgb, verbose=True)
plot_actual_vs_predicted(y_test, y_pred_xgb, "XGBoost")
===== height_cm ===== EVAR: 0.9971 MAE: 0.1862 RMSE: 0.3440 R2: 0.9971 ===== leaf_area_cm2 ===== EVAR: 0.9637 MAE: 4.1007 RMSE: 5.5724 R2: 0.9637 ===== health_score ===== EVAR: 0.9882 MAE: 0.0174 RMSE: 0.0332 R2: 0.9882
model_results["XGBoost"] = metrics_xgb
from sklearn.ensemble import RandomForestRegressor
rf = RandomForestRegressor(
n_estimators=300,
max_depth=None,
min_samples_split=2,
random_state=42,
n_jobs=-1
)
rf.fit(X_train, y_train)
y_pred_rf = rf.predict(X_test)
metrics_rf = regression_metrics(y_test, y_pred_rf, verbose=True)
plot_actual_vs_predicted(y_test, y_pred_rf, "Random Forest")
===== height_cm ===== EVAR: 0.9962 MAE: 0.1116 RMSE: 0.3948 R2: 0.9962 ===== leaf_area_cm2 ===== EVAR: 0.9620 MAE: 4.0946 RMSE: 5.6982 R2: 0.9620 ===== health_score ===== EVAR: 0.9933 MAE: 0.0119 RMSE: 0.0251 R2: 0.9933
model_results["RandomForest"] = metrics_rf
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.multioutput import MultiOutputRegressor
gbr_base = GradientBoostingRegressor(
n_estimators=300,
learning_rate=0.05,
max_depth=3,
random_state=42
)
gbr = MultiOutputRegressor(gbr_base)
gbr.fit(X_train, y_train)
y_pred_gbr = gbr.predict(X_test)
metrics_gbr = regression_metrics(y_test, y_pred_gbr, verbose=True)
plot_actual_vs_predicted(y_test, y_pred_gbr, "Gradient Boosting")
===== height_cm ===== EVAR: 0.9768 MAE: 0.6406 RMSE: 0.9758 R2: 0.9768 ===== leaf_area_cm2 ===== EVAR: 0.9445 MAE: 4.9374 RMSE: 6.8879 R2: 0.9445 ===== health_score ===== EVAR: 0.9167 MAE: 0.0502 RMSE: 0.0884 R2: 0.9166
model_results["GradientBoosting"] = metrics_gbr
from sklearn.tree import DecisionTreeRegressor
dt = DecisionTreeRegressor(
max_depth=None, # full growth
min_samples_split=2,
random_state=42
)
dt.fit(X_train, y_train)
y_pred_dt = dt.predict(X_test)
metrics_dt = regression_metrics(y_test, y_pred_dt, verbose=True)
plot_actual_vs_predicted(y_test, y_pred_dt, "Decision Tree")
===== height_cm ===== EVAR: 0.9812 MAE: 0.1532 RMSE: 0.8796 R2: 0.9811 ===== leaf_area_cm2 ===== EVAR: 0.9178 MAE: 5.3956 RMSE: 8.3782 R2: 0.9178 ===== health_score ===== EVAR: 0.9589 MAE: 0.0141 RMSE: 0.0621 R2: 0.9589
model_results["DecisionTree"] = metrics_dt
import lightgbm as lgb
from sklearn.multioutput import MultiOutputRegressor
lgb_base = lgb.LGBMRegressor(
n_estimators=300,
learning_rate=0.05,
max_depth=6,
subsample=0.9,
colsample_bytree=0.9,
random_state=42,
n_jobs=-1
)
lgb_model = MultiOutputRegressor(lgb_base)
lgb_model.fit(X_train, y_train)
y_pred_lgb = lgb_model.predict(X_test)
metrics_lgb = regression_metrics(y_test, y_pred_lgb, verbose=True)
plot_actual_vs_predicted(y_test, y_pred_lgb, "LightGBM")
[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000672 seconds. You can set `force_col_wise=true` to remove the overhead. [LightGBM] [Info] Total Bins 1546 [LightGBM] [Info] Number of data points in the train set: 11993, number of used features: 7 [LightGBM] [Info] Start training from score 14.747418 [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000582 seconds. You can set `force_col_wise=true` to remove the overhead. [LightGBM] [Info] Total Bins 1546 [LightGBM] [Info] Number of data points in the train set: 11993, number of used features: 7 [LightGBM] [Info] Start training from score 54.315238 [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000369 seconds. You can set `force_col_wise=true` to remove the overhead. [LightGBM] [Info] Total Bins 1546 [LightGBM] [Info] Number of data points in the train set: 11993, number of used features: 7 [LightGBM] [Info] Start training from score 0.264193 [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf ===== height_cm ===== EVAR: 0.9963 MAE: 0.2154 RMSE: 0.3909 R2: 0.9963 ===== leaf_area_cm2 ===== EVAR: 0.9630 MAE: 4.1109 RMSE: 5.6218 R2: 0.9630 ===== health_score ===== EVAR: 0.9876 MAE: 0.0165 RMSE: 0.0342 R2: 0.9876
model_results["LightGBM"] = metrics_lgb
from sklearn.ensemble import StackingRegressor
from sklearn.ensemble import GradientBoostingRegressor, RandomForestRegressor
from xgboost import XGBRegressor
import lightgbm as lgb
from sklearn.multioutput import MultiOutputRegressor
estimators = [
('gbr', GradientBoostingRegressor(n_estimators=300, learning_rate=0.05, max_depth=3, random_state=42)),
('lgb', lgb.LGBMRegressor(n_estimators=300, learning_rate=0.05, max_depth=6, random_state=42, n_jobs=-1))
]
stacking_model_base = StackingRegressor(
estimators=estimators,
final_estimator=GradientBoostingRegressor(n_estimators=200, learning_rate=0.05, max_depth=3, random_state=42),
n_jobs=-1,
passthrough=True
)
stacking_model = MultiOutputRegressor(stacking_model_base)
stacking_model.fit(X_train, y_train)
y_pred_stack = stacking_model.predict(X_test)
metrics_stack = regression_metrics(y_test, y_pred_stack, verbose=True)
plot_actual_vs_predicted(y_test, y_pred_stack, "Stacking Regressor")
===== height_cm ===== EVAR: 0.9960 MAE: 0.2369 RMSE: 0.4037 R2: 0.9960 ===== leaf_area_cm2 ===== EVAR: 0.9619 MAE: 4.1810 RMSE: 5.7059 R2: 0.9619 ===== health_score ===== EVAR: 0.9877 MAE: 0.0132 RMSE: 0.0340 R2: 0.9877
model_results["StackingRegressor"] = metrics_stack
import pandas as pd
def results_to_dataframe(model_results):
rows = []
for model_name, targets in model_results.items():
for target_name, metrics in targets.items():
row = {
"Model": model_name,
"Target": target_name,
"EVAR": metrics["EVAR"],
"MAE": metrics["MAE"],
"RMSE": metrics["RMSE"],
"R2": metrics["R2"]
}
rows.append(row)
return pd.DataFrame(rows)
results_df = results_to_dataframe(model_results)
results_df
| Model | Target | EVAR | MAE | RMSE | R2 | |
|---|---|---|---|---|---|---|
| 0 | XGBoost | height_cm | 0.997116 | 0.186157 | 0.344022 | 0.997116 |
| 1 | XGBoost | leaf_area_cm2 | 0.963671 | 4.100737 | 5.572361 | 0.963655 |
| 2 | XGBoost | health_score | 0.988236 | 0.017361 | 0.033235 | 0.988228 |
| 3 | RandomForest | height_cm | 0.996202 | 0.111580 | 0.394814 | 0.996202 |
| 4 | RandomForest | leaf_area_cm2 | 0.961995 | 4.094585 | 5.698208 | 0.961995 |
| 5 | RandomForest | health_score | 0.993270 | 0.011899 | 0.025132 | 0.993268 |
| 6 | GradientBoosting | height_cm | 0.976807 | 0.640644 | 0.975848 | 0.976797 |
| 7 | GradientBoosting | leaf_area_cm2 | 0.944527 | 4.937406 | 6.887915 | 0.944469 |
| 8 | GradientBoosting | health_score | 0.916674 | 0.050182 | 0.088441 | 0.916640 |
| 9 | DecisionTree | height_cm | 0.981166 | 0.153185 | 0.879646 | 0.981146 |
| 10 | DecisionTree | leaf_area_cm2 | 0.917841 | 5.395609 | 8.378197 | 0.917839 |
| 11 | DecisionTree | health_score | 0.958929 | 0.014121 | 0.062090 | 0.958914 |
| 12 | LightGBM | height_cm | 0.996277 | 0.215443 | 0.390871 | 0.996277 |
| 13 | LightGBM | leaf_area_cm2 | 0.963024 | 4.110885 | 5.621820 | 0.963007 |
| 14 | LightGBM | health_score | 0.987562 | 0.016455 | 0.034167 | 0.987559 |
| 15 | StackingRegressor | height_cm | 0.996037 | 0.236927 | 0.403688 | 0.996029 |
| 16 | StackingRegressor | leaf_area_cm2 | 0.961937 | 4.180993 | 5.705871 | 0.961893 |
| 17 | StackingRegressor | health_score | 0.987687 | 0.013176 | 0.033991 | 0.987687 |
import os
import joblib
os.makedirs("Models", exist_ok=True)
joblib.dump(stacking_model, "Models/growth_pred_model.sav")
['Models/growth_pred_model.sav']
import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style="whitegrid")
df_melt = results_df.melt(
id_vars=["Model", "Target"],
value_vars=["EVAR", "MAE", "RMSE", "R2"],
var_name="Metric",
value_name="Value"
)
plt.figure(figsize=(10, 6))
ax = sns.barplot(
data=df_melt[df_melt["Metric"]=="R2"],
x="Target",
y="Value",
hue="Model"
)
plt.title("R² Comparison Across Models and Targets", fontsize=14)
plt.ylim(0, 1.05)
plt.ylabel("R² Score")
plt.xlabel("Target Variable")
for p in ax.patches:
height = p.get_height()
ax.annotate(f"{height:.3f}",
(p.get_x() + p.get_width() / 2., height),
ha='center', va='bottom', fontsize=9, color='black', rotation=0)
plt.legend(title="Model", bbox_to_anchor=(1.05, 1), loc="upper left")
plt.tight_layout()
plt.show()
g = sns.catplot(
data=df_melt,
x="Target",
y="Value",
hue="Model",
col="Metric",
kind="bar",
height=5,
aspect=0.9
)
for ax in g.axes.flat:
for p in ax.patches:
height = p.get_height()
ax.annotate(f"{height:.3f}",
(p.get_x() + p.get_width() / 2., height),
ha='center', va='bottom', fontsize=8, rotation=0)
g.fig.suptitle("Model Performance Comparison Across Metrics and Targets", fontsize=16, y=1.02)
plt.show()
